Revision e2093926a098a8ccf0f1d10f6df8dad452cb28d3 authored by Ross Zwisler on 02 June 2017, 21:46:37 UTC, committed by Linus Torvalds on 02 June 2017, 22:07:37 UTC
We currently have two related PMD vs PTE races in the DAX code. These can both be easily triggered by having two threads reading and writing simultaneously to the same private mapping, with the key being that private mapping reads can be handled with PMDs but private mapping writes are always handled with PTEs so that we can COW. Here is the first race: CPU 0 CPU 1 (private mapping write) __handle_mm_fault() create_huge_pmd() - FALLBACK handle_pte_fault() passes check for pmd_devmap() (private mapping read) __handle_mm_fault() create_huge_pmd() dax_iomap_pmd_fault() inserts PMD dax_iomap_pte_fault() does a PTE fault, but we already have a DAX PMD installed in our page tables at this spot. Here's the second race: CPU 0 CPU 1 (private mapping read) __handle_mm_fault() passes check for pmd_none() create_huge_pmd() dax_iomap_pmd_fault() inserts PMD (private mapping write) __handle_mm_fault() create_huge_pmd() - FALLBACK (private mapping read) __handle_mm_fault() passes check for pmd_none() create_huge_pmd() handle_pte_fault() dax_iomap_pte_fault() inserts PTE dax_iomap_pmd_fault() inserts PMD, but we already have a PTE at this spot. The core of the issue is that while there is isolation between faults to the same range in the DAX fault handlers via our DAX entry locking, there is no isolation between faults in the code in mm/memory.c. This means for instance that this code in __handle_mm_fault() can run: if (pmd_none(*vmf.pmd) && transparent_hugepage_enabled(vma)) { ret = create_huge_pmd(&vmf); But by the time we actually get to run the fault handler called by create_huge_pmd(), the PMD is no longer pmd_none() because a racing PTE fault has installed a normal PMD here as a parent. This is the cause of the 2nd race. The first race is similar - there is the following check in handle_pte_fault(): } else { /* See comment in pte_alloc_one_map() */ if (pmd_devmap(*vmf->pmd) || pmd_trans_unstable(vmf->pmd)) return 0; So if a pmd_devmap() PMD (a DAX PMD) has been installed at vmf->pmd, we will bail and retry the fault. This is correct, but there is nothing preventing the PMD from being installed after this check but before we actually get to the DAX PTE fault handlers. In my testing these races result in the following types of errors: BUG: Bad rss-counter state mm:ffff8800a817d280 idx:1 val:1 BUG: non-zero nr_ptes on freeing mm: 15 Fix this issue by having the DAX fault handlers verify that it is safe to continue their fault after they have taken an entry lock to block other racing faults. [ross.zwisler@linux.intel.com: improve fix for colliding PMD & PTE entries] Link: http://lkml.kernel.org/r/20170526195932.32178-1-ross.zwisler@linux.intel.com Link: http://lkml.kernel.org/r/20170522215749.23516-2-ross.zwisler@linux.intel.com Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Reported-by: Pawel Lebioda <pawel.lebioda@intel.com> Reviewed-by: Jan Kara <jack@suse.cz> Cc: "Darrick J. Wong" <darrick.wong@oracle.com> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@lst.de> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Matthew Wilcox <mawilcox@microsoft.com> Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com> Cc: Pawel Lebioda <pawel.lebioda@intel.com> Cc: Dave Jiang <dave.jiang@intel.com> Cc: Xiong Zhou <xzhou@redhat.com> Cc: Eryu Guan <eguan@redhat.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1 parent d0f0931
pata_pcmcia.c
/*
* pata_pcmcia.c - PCMCIA PATA controller driver.
* Copyright 2005-2006 Red Hat Inc, all rights reserved.
* PCMCIA ident update Copyright 2006 Marcin Juszkiewicz
* <openembedded@hrw.one.pl>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2, or (at your option)
* any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; see the file COPYING. If not, write to
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*
* Heavily based upon ide-cs.c
* The initial developer of the original code is David A. Hinds
* <dahinds@users.sourceforge.net>. Portions created by David A. Hinds
* are Copyright (C) 1999 David A. Hinds. All Rights Reserved.
*/
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/blkdev.h>
#include <linux/delay.h>
#include <linux/slab.h>
#include <scsi/scsi_host.h>
#include <linux/ata.h>
#include <linux/libata.h>
#include <pcmcia/cistpl.h>
#include <pcmcia/ds.h>
#include <pcmcia/cisreg.h>
#include <pcmcia/ciscode.h>
#define DRV_NAME "pata_pcmcia"
#define DRV_VERSION "0.3.5"
/**
* pcmcia_set_mode - PCMCIA specific mode setup
* @link: link
* @r_failed_dev: Return pointer for failed device
*
* Perform the tuning and setup of the devices and timings, which
* for PCMCIA is the same as any other controller. We wrap it however
* as we need to spot hardware with incorrect or missing master/slave
* decode, which alas is embarrassingly common in the PC world
*/
static int pcmcia_set_mode(struct ata_link *link, struct ata_device **r_failed_dev)
{
struct ata_device *master = &link->device[0];
struct ata_device *slave = &link->device[1];
if (!ata_dev_enabled(master) || !ata_dev_enabled(slave))
return ata_do_set_mode(link, r_failed_dev);
if (memcmp(master->id + ATA_ID_FW_REV, slave->id + ATA_ID_FW_REV,
ATA_ID_FW_REV_LEN + ATA_ID_PROD_LEN) == 0) {
/* Suspicious match, but could be two cards from
the same vendor - check serial */
if (memcmp(master->id + ATA_ID_SERNO, slave->id + ATA_ID_SERNO,
ATA_ID_SERNO_LEN) == 0 && master->id[ATA_ID_SERNO] >> 8) {
ata_dev_warn(slave, "is a ghost device, ignoring\n");
ata_dev_disable(slave);
}
}
return ata_do_set_mode(link, r_failed_dev);
}
/**
* pcmcia_set_mode_8bit - PCMCIA specific mode setup
* @link: link
* @r_failed_dev: Return pointer for failed device
*
* For the simple emulated 8bit stuff the less we do the better.
*/
static int pcmcia_set_mode_8bit(struct ata_link *link,
struct ata_device **r_failed_dev)
{
return 0;
}
/**
* ata_data_xfer_8bit - Transfer data by 8bit PIO
* @qc: queued command
* @buf: data buffer
* @buflen: buffer length
* @rw: read/write
*
* Transfer data from/to the device data register by 8 bit PIO.
*
* LOCKING:
* Inherited from caller.
*/
static unsigned int ata_data_xfer_8bit(struct ata_queued_cmd *qc,
unsigned char *buf, unsigned int buflen, int rw)
{
struct ata_port *ap = qc->dev->link->ap;
if (rw == READ)
ioread8_rep(ap->ioaddr.data_addr, buf, buflen);
else
iowrite8_rep(ap->ioaddr.data_addr, buf, buflen);
return buflen;
}
/**
* pcmcia_8bit_drain_fifo - Stock FIFO drain logic for SFF controllers
* @qc: command
*
* Drain the FIFO and device of any stuck data following a command
* failing to complete. In some cases this is necessary before a
* reset will recover the device.
*
*/
static void pcmcia_8bit_drain_fifo(struct ata_queued_cmd *qc)
{
int count;
struct ata_port *ap;
/* We only need to flush incoming data when a command was running */
if (qc == NULL || qc->dma_dir == DMA_TO_DEVICE)
return;
ap = qc->ap;
/* Drain up to 64K of data before we give up this recovery method */
for (count = 0; (ap->ops->sff_check_status(ap) & ATA_DRQ)
&& count++ < 65536;)
ioread8(ap->ioaddr.data_addr);
if (count)
ata_port_warn(ap, "drained %d bytes to clear DRQ\n", count);
}
static struct scsi_host_template pcmcia_sht = {
ATA_PIO_SHT(DRV_NAME),
};
static struct ata_port_operations pcmcia_port_ops = {
.inherits = &ata_sff_port_ops,
.sff_data_xfer = ata_sff_data_xfer_noirq,
.cable_detect = ata_cable_40wire,
.set_mode = pcmcia_set_mode,
};
static struct ata_port_operations pcmcia_8bit_port_ops = {
.inherits = &ata_sff_port_ops,
.sff_data_xfer = ata_data_xfer_8bit,
.cable_detect = ata_cable_40wire,
.set_mode = pcmcia_set_mode_8bit,
.sff_drain_fifo = pcmcia_8bit_drain_fifo,
};
static int pcmcia_check_one_config(struct pcmcia_device *pdev, void *priv_data)
{
int *is_kme = priv_data;
if ((pdev->resource[0]->flags & IO_DATA_PATH_WIDTH)
!= IO_DATA_PATH_WIDTH_8) {
pdev->resource[0]->flags &= ~IO_DATA_PATH_WIDTH;
pdev->resource[0]->flags |= IO_DATA_PATH_WIDTH_AUTO;
}
pdev->resource[1]->flags &= ~IO_DATA_PATH_WIDTH;
pdev->resource[1]->flags |= IO_DATA_PATH_WIDTH_8;
if (pdev->resource[1]->end) {
pdev->resource[0]->end = 8;
pdev->resource[1]->end = (*is_kme) ? 2 : 1;
} else {
if (pdev->resource[0]->end < 16)
return -ENODEV;
}
return pcmcia_request_io(pdev);
}
/**
* pcmcia_init_one - attach a PCMCIA interface
* @pdev: pcmcia device
*
* Register a PCMCIA IDE interface. Such interfaces are PIO 0 and
* shared IRQ.
*/
static int pcmcia_init_one(struct pcmcia_device *pdev)
{
struct ata_host *host;
struct ata_port *ap;
int is_kme = 0, ret = -ENOMEM, p;
unsigned long io_base, ctl_base;
void __iomem *io_addr, *ctl_addr;
int n_ports = 1;
struct ata_port_operations *ops = &pcmcia_port_ops;
/* Set up attributes in order to probe card and get resources */
pdev->config_flags |= CONF_ENABLE_IRQ | CONF_AUTO_SET_IO |
CONF_AUTO_SET_VPP | CONF_AUTO_CHECK_VCC;
/* See if we have a manufacturer identifier. Use it to set is_kme for
vendor quirks */
is_kme = ((pdev->manf_id == MANFID_KME) &&
((pdev->card_id == PRODID_KME_KXLC005_A) ||
(pdev->card_id == PRODID_KME_KXLC005_B)));
if (pcmcia_loop_config(pdev, pcmcia_check_one_config, &is_kme)) {
pdev->config_flags &= ~CONF_AUTO_CHECK_VCC;
if (pcmcia_loop_config(pdev, pcmcia_check_one_config, &is_kme))
goto failed; /* No suitable config found */
}
io_base = pdev->resource[0]->start;
if (pdev->resource[1]->end)
ctl_base = pdev->resource[1]->start;
else
ctl_base = pdev->resource[0]->start + 0x0e;
if (!pdev->irq)
goto failed;
ret = pcmcia_enable_device(pdev);
if (ret)
goto failed;
/* iomap */
ret = -ENOMEM;
io_addr = devm_ioport_map(&pdev->dev, io_base, 8);
ctl_addr = devm_ioport_map(&pdev->dev, ctl_base, 1);
if (!io_addr || !ctl_addr)
goto failed;
/* Success. Disable the IRQ nIEN line, do quirks */
iowrite8(0x02, ctl_addr);
if (is_kme)
iowrite8(0x81, ctl_addr + 0x01);
/* FIXME: Could be more ports at base + 0x10 but we only deal with
one right now */
if (resource_size(pdev->resource[0]) >= 0x20)
n_ports = 2;
if (pdev->manf_id == 0x0097 && pdev->card_id == 0x1620)
ops = &pcmcia_8bit_port_ops;
/*
* Having done the PCMCIA plumbing the ATA side is relatively
* sane.
*/
ret = -ENOMEM;
host = ata_host_alloc(&pdev->dev, n_ports);
if (!host)
goto failed;
for (p = 0; p < n_ports; p++) {
ap = host->ports[p];
ap->ops = ops;
ap->pio_mask = ATA_PIO0; /* ISA so PIO 0 cycles */
ap->flags |= ATA_FLAG_SLAVE_POSS;
ap->ioaddr.cmd_addr = io_addr + 0x10 * p;
ap->ioaddr.altstatus_addr = ctl_addr + 0x10 * p;
ap->ioaddr.ctl_addr = ctl_addr + 0x10 * p;
ata_sff_std_ports(&ap->ioaddr);
ata_port_desc(ap, "cmd 0x%lx ctl 0x%lx", io_base, ctl_base);
}
/* activate */
ret = ata_host_activate(host, pdev->irq, ata_sff_interrupt,
IRQF_SHARED, &pcmcia_sht);
if (ret)
goto failed;
pdev->priv = host;
return 0;
failed:
pcmcia_disable_device(pdev);
return ret;
}
/**
* pcmcia_remove_one - unplug an pcmcia interface
* @pdev: pcmcia device
*
* A PCMCIA ATA device has been unplugged. Perform the needed
* cleanup. Also called on module unload for any active devices.
*/
static void pcmcia_remove_one(struct pcmcia_device *pdev)
{
struct ata_host *host = pdev->priv;
if (host)
ata_host_detach(host);
pcmcia_disable_device(pdev);
}
static const struct pcmcia_device_id pcmcia_devices[] = {
PCMCIA_DEVICE_FUNC_ID(4),
PCMCIA_DEVICE_MANF_CARD(0x0000, 0x0000), /* Corsair */
PCMCIA_DEVICE_MANF_CARD(0x0007, 0x0000), /* Hitachi */
PCMCIA_DEVICE_MANF_CARD(0x000a, 0x0000), /* I-O Data CFA */
PCMCIA_DEVICE_MANF_CARD(0x001c, 0x0001), /* Mitsubishi CFA */
PCMCIA_DEVICE_MANF_CARD(0x0032, 0x0704),
PCMCIA_DEVICE_MANF_CARD(0x0032, 0x2904),
PCMCIA_DEVICE_MANF_CARD(0x0045, 0x0401), /* SanDisk CFA */
PCMCIA_DEVICE_MANF_CARD(0x004f, 0x0000), /* Kingston */
PCMCIA_DEVICE_MANF_CARD(0x0097, 0x1620), /* TI emulated */
PCMCIA_DEVICE_MANF_CARD(0x0098, 0x0000), /* Toshiba */
PCMCIA_DEVICE_MANF_CARD(0x00a4, 0x002d),
PCMCIA_DEVICE_MANF_CARD(0x00ce, 0x0000), /* Samsung */
PCMCIA_DEVICE_MANF_CARD(0x0319, 0x0000), /* Hitachi */
PCMCIA_DEVICE_MANF_CARD(0x2080, 0x0001),
PCMCIA_DEVICE_MANF_CARD(0x4e01, 0x0100), /* Viking CFA */
PCMCIA_DEVICE_MANF_CARD(0x4e01, 0x0200), /* Lexar, Viking CFA */
PCMCIA_DEVICE_PROD_ID123("Caravelle", "PSC-IDE ", "PSC000", 0x8c36137c, 0xd0693ab8, 0x2768a9f0),
PCMCIA_DEVICE_PROD_ID123("CDROM", "IDE", "MCD-601p", 0x1b9179ca, 0xede88951, 0x0d902f74),
PCMCIA_DEVICE_PROD_ID123("PCMCIA", "IDE CARD", "F1", 0x281f1c5d, 0x1907960c, 0xf7fde8b9),
PCMCIA_DEVICE_PROD_ID12("ARGOSY", "CD-ROM", 0x78f308dc, 0x66536591),
PCMCIA_DEVICE_PROD_ID12("ARGOSY", "PnPIDE", 0x78f308dc, 0x0c694728),
PCMCIA_DEVICE_PROD_ID12("CNF ", "CD-ROM", 0x46d7db81, 0x66536591),
PCMCIA_DEVICE_PROD_ID12("CNF CD-M", "CD-ROM", 0x7d93b852, 0x66536591),
PCMCIA_DEVICE_PROD_ID12("Creative Technology Ltd.", "PCMCIA CD-ROM Interface Card", 0xff8c8a45, 0xfe8020c4),
PCMCIA_DEVICE_PROD_ID12("Digital Equipment Corporation.", "Digital Mobile Media CD-ROM", 0x17692a66, 0xef1dcbde),
PCMCIA_DEVICE_PROD_ID12("EXP", "CD+GAME", 0x6f58c983, 0x63c13aaf),
PCMCIA_DEVICE_PROD_ID12("EXP ", "CD-ROM", 0x0a5c52fd, 0x66536591),
PCMCIA_DEVICE_PROD_ID12("EXP ", "PnPIDE", 0x0a5c52fd, 0x0c694728),
PCMCIA_DEVICE_PROD_ID12("FREECOM", "PCCARD-IDE", 0x5714cbf7, 0x48e0ab8e),
PCMCIA_DEVICE_PROD_ID12("HITACHI", "FLASH", 0xf4f43949, 0x9eb86aae),
PCMCIA_DEVICE_PROD_ID12("HITACHI", "microdrive", 0xf4f43949, 0xa6d76178),
PCMCIA_DEVICE_PROD_ID12("Hyperstone", "Model1", 0x3d5b9ef5, 0xca6ab420),
PCMCIA_DEVICE_PROD_ID12("IBM", "microdrive", 0xb569a6e5, 0xa6d76178),
PCMCIA_DEVICE_PROD_ID12("IBM", "IBM17JSSFP20", 0xb569a6e5, 0xf2508753),
PCMCIA_DEVICE_PROD_ID12("KINGSTON", "CF CARD 1GB", 0x2e6d1829, 0x55d5bffb),
PCMCIA_DEVICE_PROD_ID12("KINGSTON", "CF CARD 4GB", 0x2e6d1829, 0x531e7d10),
PCMCIA_DEVICE_PROD_ID12("KINGSTON", "CF8GB", 0x2e6d1829, 0xacbe682e),
PCMCIA_DEVICE_PROD_ID12("IO DATA", "CBIDE2 ", 0x547e66dc, 0x8671043b),
PCMCIA_DEVICE_PROD_ID12("IO DATA", "PCIDE", 0x547e66dc, 0x5c5ab149),
PCMCIA_DEVICE_PROD_ID12("IO DATA", "PCIDEII", 0x547e66dc, 0xb3662674),
PCMCIA_DEVICE_PROD_ID12("LOOKMEET", "CBIDE2 ", 0xe37be2b5, 0x8671043b),
PCMCIA_DEVICE_PROD_ID12("M-Systems", "CF300", 0x7ed2ad87, 0x7e9e78ee),
PCMCIA_DEVICE_PROD_ID12("M-Systems", "CF500", 0x7ed2ad87, 0x7a13045c),
PCMCIA_DEVICE_PROD_ID2("NinjaATA-", 0xebe0bd79),
PCMCIA_DEVICE_PROD_ID12("PCMCIA", "CD-ROM", 0x281f1c5d, 0x66536591),
PCMCIA_DEVICE_PROD_ID12("PCMCIA", "PnPIDE", 0x281f1c5d, 0x0c694728),
PCMCIA_DEVICE_PROD_ID12("SHUTTLE TECHNOLOGY LTD.", "PCCARD-IDE/ATAPI Adapter", 0x4a3f0ba0, 0x322560e1),
PCMCIA_DEVICE_PROD_ID12("SEAGATE", "ST1", 0x87c1b330, 0xe1f30883),
PCMCIA_DEVICE_PROD_ID12("SAMSUNG", "04/05/06", 0x43d74cb4, 0x6a22777d),
PCMCIA_DEVICE_PROD_ID12("SMI VENDOR", "SMI PRODUCT", 0x30896c92, 0x703cc5f6),
PCMCIA_DEVICE_PROD_ID12("TOSHIBA", "MK2001MPL", 0xb4585a1a, 0x3489e003),
PCMCIA_DEVICE_PROD_ID1("TRANSCEND 512M ", 0xd0909443),
PCMCIA_DEVICE_PROD_ID12("TRANSCEND", "TS1GCF45", 0x709b1bf1, 0xf68b6f32),
PCMCIA_DEVICE_PROD_ID12("TRANSCEND", "TS1GCF80", 0x709b1bf1, 0x2a54d4b1),
PCMCIA_DEVICE_PROD_ID12("TRANSCEND", "TS2GCF120", 0x709b1bf1, 0x969aa4f2),
PCMCIA_DEVICE_PROD_ID12("TRANSCEND", "TS4GCF120", 0x709b1bf1, 0xf54a91c8),
PCMCIA_DEVICE_PROD_ID12("TRANSCEND", "TS4GCF133", 0x709b1bf1, 0x7558f133),
PCMCIA_DEVICE_PROD_ID12("TRANSCEND", "TS8GCF133", 0x709b1bf1, 0xb2f89b47),
PCMCIA_DEVICE_PROD_ID12("WIT", "IDE16", 0x244e5994, 0x3e232852),
PCMCIA_DEVICE_PROD_ID12("WEIDA", "TWTTI", 0xcc7cf69c, 0x212bb918),
PCMCIA_DEVICE_PROD_ID1("STI Flash", 0xe4a13209),
PCMCIA_DEVICE_PROD_ID12("STI", "Flash 5.0", 0xbf2df18d, 0x8cb57a0e),
PCMCIA_MFC_DEVICE_PROD_ID12(1, "SanDisk", "ConnectPlus", 0x7a954bd9, 0x74be00c6),
PCMCIA_DEVICE_PROD_ID2("Flash Card", 0x5a362506),
PCMCIA_DEVICE_NULL,
};
MODULE_DEVICE_TABLE(pcmcia, pcmcia_devices);
static struct pcmcia_driver pcmcia_driver = {
.owner = THIS_MODULE,
.name = DRV_NAME,
.id_table = pcmcia_devices,
.probe = pcmcia_init_one,
.remove = pcmcia_remove_one,
};
module_pcmcia_driver(pcmcia_driver);
MODULE_AUTHOR("Alan Cox");
MODULE_DESCRIPTION("low-level driver for PCMCIA ATA");
MODULE_LICENSE("GPL");
MODULE_VERSION(DRV_VERSION);
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...